Review
Therapeutic intervention with complement and β-glucan in cancer

https://doi.org/10.1016/S0162-3109(99)00013-2Get rights and content

Abstract

Complement (C) has two major effector systems available for host defense. The membrane attack complex (MAC) generated from components C5–C9 can form membrane-penetrating lesions that lead to cell death by causing a rapid loss of cytoplasmic components. The MAC is only effective against pathogens with outer phospholipid membranes, and cannot kill Gram-positive bacteria or yeast whose membranes are protected by cell walls. The most important effector mechanism of C is the opsonization of microbial pathogens with the serum protein C3 that leads to their high avidity attachment to the C3-receptors of phagocytic cells. Pathogens that activate complement are first coated with the C3b fragment of C3, which is rapidly proteolyzed into the iC3b fragment by serum factor I. These iC3b fragments serve to promote the high avidity attachment of the `iC3b-opsonized' pathogens to the iC3b-receptors (CR3, CD11b/CD18) of phagocytic cells and natural killer (NK) cells, stimulating phagocytosis and/or cytotoxic degranulation. Host cells, including neoplastic tumor cells, have been endowed with natural mechanisms for self-protection against both the MAC and the cytotoxic activation of CR3. This review discusses a novel type of immunotherapy for cancer that uses soluble yeast β-glucan to override the normal resistance of iC3b-opsonized tumor cells to the cytotoxic activation of phagocyte and NK cell CR3, allowing this important effector mechanism of the C system to function against tumor cells in the same way that it normally functions against bacteria and yeast. Moreover, the cytotoxic activation of β-glucan-primed NK cell CR3 by iC3b-opsonized tumors is shown to be accompanied by a tumor-localized secretion of the cytokines TNFα, IFNα, IFNγ, and IL-6.

Introduction

The field of tumor immunology has had a checkered history where periods of great enthusiasm were dashed by failures in the clinic. In the modern era, tumor Ags were dismissed as fetal Ags, and the hope of rIFNγ and rIL-2 was met with toxicity and minimal success. Some of these past problems can be ascribed to the complexity of the immune system, and the remaining, to a heterogeneity of the malignant process both among different patients and in the same patient at different stages of their disease. An increasing number of tumor-specific Ags have been identified and we now have a better understanding of Ag presentation and the pathways for generating humoral vs. cellular immunity. An increasing awareness has occurred that the immune destruction of tumors requires a combination of effector mechanisms, and that a single vaccine, cytokine, or biological response modifier (BRM) is unlikely to be successful in the majority of patients. For example, vaccines may elicit immune cytotoxic T lymphocyte (CTL) cells and/or humoral Ab responses and yet both have shortcomings. Antibodies are frequently ineffective because normal host cell proteins (DAF, MCP, and CD59) inhibit complement (C)-mediated cytotoxicity (Kojima et al., 1993; Varsano et al., 1998; Venneker et al., 1998), and iC3b-opsonization of tumors does not recruit phagocytes and natural killer (NK) cells. Antibody-dependent cell-mediated immunity (ADCC) is thought to fail because the IgG density achieved on tumors is too low and FcγRIII-mediated cytotoxicity is suppressed by NK cell recognition of tumor cell MHC class I (Binstadt et al., 1996). In recent years, it was widely believed that cellular immunity could succeed where humoral immunity had failed. However, the identification of peptide epitopes that can be presented by all types of HLA molecules has proved a daunting task, and this effort could be futile if most tumors lose HLA class I as part of the metastatic process (Cordon-Cardo et al., 1991; Garrido et al., 1993; Esteban et al., 1996). A good CTL response could even function to select for class I-negative tumor cells (Khanna, 1998). A recent study of metastatic mammary carcinoma and melanoma reported that tumors from >70% of patients no longer expressed class I, and therefore, a CTL-targeted vaccine was doomed to fail in patients with breast cancer or melanoma (Porgador et al., 1997). As this study proposed, NK cells may be particularly effective against tumors that lose class I because NK cell cytotoxicity is suppressed by recognition of class I (Lanier, 1998; López-Botet et al., 1998). However, rIL-2 therapy that produces activated NK cells (i.e., LAK cells) does not specifically target the NK cells to tumors and has several toxic side effects (Kammula et al., 1998; Whiteside et al., 1998).

The iC3b-receptor, CR3, known also as Mac-1 or αMβ2-integrin, has two major functions. As the Mac-1 adhesion molecule, it mediates the diapedesis of leukocytes through the endothelium via generation of a high-affinity binding site for ICAM-1 (Springer, 1994; Hogg and Berlin, 1995; Sugimori et al., 1997). As CR3, it stimulates phagocytosis and degranulation in response to microorganisms or immune complexes opsonized with iC3b (Petty and Todd, 1993; Ross and Větvička, 1993; Sutterwala et al., 1996). For these functions, the Mac-1/CR3 molecule goes through a series of `inside-out' and/or `outside-in' signaling steps that result in exposure of high-affinity binding sites and/or an altered linkage to the actin cytoskeleton (Brown and Hogg, 1996; Newton et al., 1997). The nature of these activation and signaling pathways has not been completely defined, and it is particularly unknown whether activation for cytotoxic responses involves a similar pathway of events as the signaling for acquisition of the high-affinity ICAM-1 binding site. Our research has focused on the cytotoxic functions of CR3 whereas the majority of research by other investigators on Mac-1 has focused on mechanisms for development of its adhesion functions.

In 1987, it was shown that neutrophil CR3-dependent phagocytosis or degranulation in response to iC3b-opsonized yeast required ligation of two distinct binding sites in CR3, one for iC3b and a second site for β-glucan (Cain et al., 1987; Ross et al., 1987). Subsequent research mapped each of these binding sites to domains within the α-chain of CR3, CD11b (Fig. 1). All protein ligands of CR3, including iC3b, bind to overlapping sites contained within the I-domain of CD11b (Diamond et al., 1993, Diamond et al., 1995; Ueda et al., 1994; Zhou et al., 1994; Zhang and Plow, 1996; Balsam et al., 1998). Using flow cytometry with fluorescein isothiocyanate (FITC)-labeled polysaccharides and CHO cells expressing recombinant chimeras between CD11b and CD11c, the lectin site was mapped to a region of CD11b located C-terminal to the I-domain (Thornton et al., 1996). The lectin domain functions to prime CR3 for cytotoxic responses (Větvička et al., 1996). C3-opsonized microorganisms present iC3b in combination with cell wall polysaccharides, such that both of these domains of CR3 become attached to microbial pathogens, stimulating phagocytosis and cytotoxic degranulation (Fig. 2) (Cain et al., 1987). NK cell CR3 functions in a similar manner as phagocyte CR3 in mediating cytotoxic degranulation in response to dual ligation of these two CR3 binding sites. This is the mechanism used by NK cells for CR3-dependent cytotoxicity of Candida albicans (Forsyth and Mathews, 1996). The lack of similar CR3-binding polysaccharides on human cells explains the inability of CR3 to mediate phagocytosis or extracellular cytotoxicity of erythrocytes or tumor cells opsonized with iC3b (Perlmann et al., 1975; Newman and Johnston, 1979; Schreiber et al., 1982; Wright and Silverstein, 1982; Wright et al., 1983; Wright, 1985). Host cell membranes opsonized with iC3b engage only the I-domain of CD11b and not the lectin site. On the other hand, small soluble β(1–3)-glucan polysaccharides isolated from fungi can bind to the lectin site of CR3 with high affinity and prime the receptor for subsequent cytotoxic activation by iC3b-tumor cells that are otherwise inert in stimulating CR3-dependent cytotoxicity (Větvička et al., 1996, Větvička et al., 1997). Polysaccharide priming of CR3 involves a Mg2+ and protein tyrosine kinase (PTK)-dependent conformational change in CD11b that exposes the activation epitope defined by mAb CBRM1/5, but not the high-affinity ICAM-1 reporter epitope defined by mAb 24 (Větvička et al., 1996). In addition to serving as a receptor for exogenous polysaccharides on microorganisms, the lectin site also functions to link endogenous neutrophil membrane glycoproteins to CR3. This linkage to CR3 occurs with a large family of neutrophil membrane glycoprotein receptors bearing a phosphatidylinositol glycolipid (PIG) anchor (e.g., CD14, CD16, CD59, and CD87), and the linkage to CR3 via the lectin site provides a mechanism for transmembrane signaling to receptors that have no transmembrane signaling ability of their own (nor any connection to the actin cytoskeleton for mediating particle ingestion or adhesion). For example, FcγRIIIB (CD16) binds to the lectin site of neighboring CR3, thereby acquiring through the lectin site of CR3 the ability to stimulate phagocytosis or degranulation (Zhou et al., 1993; Krauss et al., 1994; Poo et al., 1995). Likewise, CD14 is able to mediate phagocytosis of Escherichia coli through attachment first to the bacteria via lipopolysaccharide (LPS), and then secondarily following CD14 attachment to the lectin site of CR3, the CD14/CR3 membrane complex is able to mediate the ingestion of E. coli (Zarewych et al., 1996; Ingalls et al., 1998). Recent studies have shown that CD59 and CD87 (uPAR) use the lectin site of CR3 to mediate adhesion rather than cytotoxicity. Not only is adhesion prevented or disrupted by oligosaccharides that compete with the lectin site (Gyetko et al., 1995; Sitrin et al., 1996; Cramer et al., 1998), but also CD87 knock-out mice are unable to generate the high-affinity ICAM-1 binding site in the I-domain of either CD11a or CD11b (May et al., 1998). Thus, the lectin site is not only essential for CR3-mediated cytotoxicity but also is apparently required for Mac-1-dependent adhesion. This finding, in combination with other data showing similar cell surface lectin-dependent complexes involving LFA-1 or CR4 (CD11c), suggests that homologous lectin sites may be present in the other CD11 family members and participate in similar lectin–carbohydrate complexes needed for transmembrane signaling functions (Petty and Todd, 1996; Todd and Petty, 1997).

Structural analyses of rCD11b I-domains using X-ray crystallography and mutagenesis have proposed a `metal ion-dependent adhesion site' (MIDAS) with central Mg2+ whose structure allows key residues to be exposed and/or reoriented to provide binding sites of varying affinity for the protein ligands used for Mac-1 adhesion (Kamata et al., 1995; Lee et al., 1995a, Lee et al., 1995b; Rieu et al., 1996; Zhang and Plow, 1997). The N-terminal domain folds back onto the divalent cation binding region, forming a loop termed as β-propeller (Fig. 1) (Lu et al., 1998). The functional contribution of regions outside the I-domain is only beginning to be explored. The C-terminal location of the lectin site of CD11b was recently confirmed in a study of rCR3 binding to C. albicans that suggested that ligation of Candida polysaccharides to the lectin site caused an increased affinity of a second binding site for Candida located in the I-domain (Forsyth et al., 1998). Our studies have suggested a site, located C-terminal to both the I-domain and the divalent-cation binding repeats sequence, that became covered or hidden when mAbs were attached to distal sites in the I-domain (Thornton et al., 1996). A similar finding of lectin site blockade by a mAb to the I-domain was recently also made with mouse CR3 (Xia et al., 1999). Other studies with rCD11b expressed without CD18 by insect cells infected with recombinant baculoviruses showed that the binding to β-glucan-FITC or 125I-β-glucan to rCD11b could be blocked by mAbs to the I-domain, as well as by mAbs to C-terminal domain epitopes (Xia and Ross, 1998). However, most importantly, rCD11b fragments from which the I-domain had been deleted retained lectin site activity, and this activity was blocked only by mAbs to C-terminal epitopes and not by mAbs to the I-domain. From these data, it was deduced that the lectin site was formed entirely by CD11b, and that lectin site exposure on CD11b did not require the CD11b/CD18 heterodimer.

These data suggest that occupation of the lectin site by a glycoprotein such as CD59 or CD87 can stimulate a change in the conformation of the distal I-domain (such as an increased affinity of the MIDAS for ICAM-1), and conversely, occupation of the I-domain by a mAb can change the conformation of the distal lectin site such that its binding site for soluble polysaccharides is no longer exposed. On the other hand, occupation of the lectin site by a soluble polysaccharide appears to inhibit complex formation with CD59 or CD87, thereby preventing development of the high-affinity binding site for ICAM-1, but priming CR3 for cytotoxic activation in response to ligation of the I-domain to an iC3b-opsonized target cell.

Biological response modifiers derived from microbial products have represented important tools for defining mechanisms of host defense. However, most BRMs have remained classified as non-specific because their exact mode of action was unknown. β-Glucan BRM were first reported 35 years ago and have been extensively investigated for both their anti-tumor and anti-infective activity. Most β-glucan BRMs are derived from yeast or fungi and have a backbone structure of linear β-1,3-linked d-glucose molecules (β-1,3-d-glucan) with β-1,6-linked side chains of β-1,3-d-glucan of varying sizes that occur at different intervals along the backbone (Bohn and BeMiller, 1995; Misaki and Kakuta, 1997). The frequency of the β-1,6-linked side chains, known as the degree of substitution or branching frequency, regulates secondary structure, solubility (Ohno et al., 1986; Maeda et al., 1988; Saito et al., 1991), and ultimately, the affinity of individual types of β-glucans for the lectin site of CR3 (Thornton et al., 1996; Ross et al., 1998a). However, our laboratory has reported the only studies that have related β-glucan receptor binding affinity to function in mediating leukocyte (neutrophil, monocyte, macrophage, NK cell) activation for tumoricidal activity or cytokine release. Over 500 papers during the past 30 years, predominantly in the Japanese pharmaceutical literature, have examined β-glucan structure only in relation to tumoricidal or bactericidal activity, and have not attempted to identify its target receptor as a way of defining optimal polysaccharide structure. These reports have shown that β-glucans, either soluble or particulate, and isolated from various natural sources, exhibit antitumor and antimicrobial activities in several animal species including mice (Diller et al., 1963; Chihara et al., 1969; Di Luzio et al., 1979; Williams et al., 1983; Ohno et al., 1984; Mimura et al., 1985; Seljelid, 1986; Kurachi et al., 1990; Kitamura et al., 1994; Sveinbjornsson et al., 1998). Some of the soluble fungal β-glucans have been applied clinically for tumor immunotherapy, such as lentinan, derived from an edible mushroom (Chihara et al., 1969), and schizophyllan (i.e., SSG or Sizofiran) isolated from the culture filtrate of Schizophyllum commune (Komatsu et al., 1969; Mansell et al., 1978; Nakao et al., 1983; Fujimoto et al., 1984; Wakui et al., 1986; Taguchi, 1987; Fujimoto, 1989; Chen and Hasumi, 1993; Tari et al., 1994; Nakano et al., 1996; Matsuoka et al., 1997). In vitro studies have shown that β-glucans activate macrophages, neutrophils, and NK cells to kill sensitive tumor cells (Cook et al., 1978).

Although somewhat controversial (Czop and Kay, 1991; Zimmerman et al., 1998), recent data suggest that CR3 serves as the major, if not the only receptor for β-glucans with human (Thornton et al., 1996) or mouse (Xia et al., 1999) leukocytes, and therefore, may be responsible for all reported functions of β-glucans in vitro and in vivo. Unlike other `non-specific' BRMs, β-glucan specifically targets macrophages, neutrophils, and NK cells to tumors that are opsonized with Ab and C3, and therefore, β-glucan has the same specificity as the tumor-opsonizing Ab. This research has particularly shown the therapeutic value in mice of small soluble β-glucans (5–20 kDa) that bind to CR3 with high affinity and prime the receptor for subsequent cytotoxic activation if, and only if, CR3 subsequently comes in contact with an iC3b-opsonized target cell. Particulate β-glucan and high molecular weight (m.w.) soluble β-glucans such as lentinan and schizophyllan (>500 kDa) that have been used for patient therapy in Japan have been shown to be large enough to cross-link membrane CR3 of neutrophils and monocytes, triggering respiratory bursts, degranulation, and cytokine release in the absence of target cells (Ross et al., 1987; Doita et al., 1991; Ohno et al., 1993; Ross and Větvička, 1996; Větvička et al., 1996; Ljungman et al., 1998). Several studies have shown the safety of soluble β-glucans and the absence of undesirable side effects (Williams et al., 1988, Williams et al., 1991). The only problems reported have occurred with high m.w. soluble or particulate β-glucans (Maeda et al., 1996; Yoshioka et al., 1998). By comparison, Betafectin, a relatively low m.w. soluble β-glucan (∼150 kDa), does not stimulate cytokine release (Bleicher and Mackin, 1995), probably because it is too small to cross-link membrane CR3.

The targets for β-glucan-primed CR3 appear to be any iC3b-opsonized host cell or microbial pathogen, and perhaps also tumor cells or parasites bearing endogenous ligands for CR3, although such CR3 ligands have only been detected on K562 cells (Větvička et al., 1996) and certain leishmania species (Russell and Wright, 1988). Tumors appear to be opsonized frequently with Ab and C3 as the result of an ineffective humoral response, and this could be enhanced therapeutically through either vaccines or mAbs to tumor Ags. Virus-infected cells or cells infected with intracellular bacteria also frequently activate C, either because they have become spontaneous activators of the alternative pathway or through Abs that activate the classical pathway of C. This common feature of target cell-bound iC3b appears to explain the wide range of diseases that respond to therapy with β-glucans. Thus, it proposed that resistance to β-glucan therapy corresponds to the absence of tumor cell- or microbe-bound iC3b, and that the success of β-glucan therapy can be enhanced by agents such as vaccines that enhance the target cell density of bound Ab and iC3b.

Section snippets

Research on human leukocytes and tumor cells in vitro

Previous reports had suggested that malignant cells frequently generated a humoral response that was ineffective in tumor destruction. Immunohistochemical staining of excised tumors for Ig and C3, as well as circulating tumor-reactive Abs, have been noted in patients with mammary carcinoma and cancers of the lung and colon (Irie et al., 1974; Seegal et al., 1976; Niculescu et al., 1992; Kotera et al., 1994). This natural humoral immune attack on tumors does not appear to prevent tumor growth,

References (116)

  • J.-O. Lee et al.

    Crystal structure of the A domain from the α subunit of integrin CR3 (CD11b/CD18)

    Cell

    (1995)
  • J.O. Lee et al.

    Two conformations of the integrin A-domain (I-domain): a pathway for activation?

    Structure

    (1995)
  • A.G. Ljungman et al.

    (1–>3)-β-d-Glucan stimulates nitric oxide generation and cytokine mRNA expression in macrophages

    Environ. Toxicol. Pharmacol.

    (1998)
  • C.F. Lu et al.

    The structure of the β-propeller domain and C-terminal region of the integrin αM subunit—dependence on β subunit association and prediction of domains

    J. Biol. Chem.

    (1998)
  • T. Nakano et al.

    Intratumoral administration of sizofiran activates Langerhans cell and T cell infiltration in cervical cancer

    Clin. Immunol. Immunopathol.

    (1996)
  • H.R. Petty et al.

    Integrins as promiscuous signal transduction devices

    Immunol. Today

    (1996)
  • H. Poo et al.

    Interaction of Fcγ receptor type IIIB with complement receptor type 3 in fibroblast transfectants: evidence from lateral diffusion and resonance energy transfer studies

    J. Mol. Biol.

    (1995)
  • P. Rieu et al.

    Solvent-accessible residues on the metal ion-dependent adhesion site face of integrin CR3 mediate its binding to the neutrophil inhibitory factor

    J. Biol. Chem.

    (1996)
  • G.D. Ross et al.

    Therapeutic intervention with complement and complement receptors in cancer

    Mol. Immunol.

    (1998)
  • H. Saito et al.

    Relationship between conformation and biological response for (1–>3)-β-d-glucans in the activation of coagulation factor G from limulus amebocyte lysate and host-mediated antitumor activity. Demonstration of single-helix conformation as a stimulant

    Carbohydr. Res.

    (1991)
  • R.D. Schreiber et al.

    The role of C3 fragments in endocytosis and extracellular cytotoxic reactions by polymorphonuclear leukocytes

    Clin. Immunol. Immunopathol.

    (1982)
  • T.A. Springer

    Traffic signals for lymphocyte recirculation and leukocyte emigration: the multistep paradigm

    Cell

    (1994)
  • T. Sugimori et al.

    Emerging paradigms of integrin ligand binding and activation

    Kidney Int.

    (1997)
  • R.F. Todd et al.

    β2(CD11/CD18) integrins can serve as signaling partners for other leukocyte receptors

    J. Lab. Clin. Med.

    (1997)
  • G.T. Venneker et al.

    Relative contributions of decay accelerating factor (DAF), membrane cofactor protein (MCP) and CD59 in the protection of melanocytes from homologous complement

    Immunobiology

    (1998)
  • B. Au et al.

    Zymosan-induced IL-8 release from human neutrophils involves activation via the CD11b/CD18 receptor and endogenous platelet-activating factor as an autocrine modulator

    J. Immunol.

    (1994)
  • L.B. Balsam et al.

    Functional mapping of CD11b/CD18 epitopes important in neutrophil–epithelial interactions: a central role of the I domain

    J. Immunol.

    (1998)
  • P. Bleicher et al.

    Betafectin PPG-glucan: a novel carbohydrate immunomodulator with anti-infective properties

    J. Biotechnol. Healthcare

    (1995)
  • J.A. Cain et al.

    Role of complement receptor type three and serum opsonins in the neutrophil response to yeast

    Complement. Inflamm.

    (1987)
  • J.-T. Chen et al.

    Activation of peritoneal macrophages in patients with gynecological malignancies by sizofiran and recombinant interferon-γ

    Biotherapy

    (1993)
  • G. Chihara et al.

    Inhibition of mouse sarcoma 180 by polysaccharides from Lentinus edodes (Berk.)

    Sing. Nat.

    (1969)
  • Cook, J.A., Taylor, D., Cohen, C., Rodrique, J., Malshet, V., Di Luzio, N.R., 1978. Comparative evaluation of the role...
  • C. Cordon-Cardo et al.

    Expression of HLA-A,B,C antigens on primary and metastatic tumor cell populations of human carcinomas

    Cancer Res.

    (1991)
  • J.K. Czop et al.

    Isolation and characterization of β-glucan receptors on human mononuclear phagocytes

    J. Exp. Med.

    (1991)
  • G. Dennert et al.

    Antitumor polysaccharide lentinan. A T cell adjuvant

    Journal of the National Cancer Institute

    (1973)
  • M.S. Diamond et al.

    The I domain is a major recognition site on the leukocyte integrin Mac-1 (CD11b/CD18) for four distinct adhesion ligands

    J. Cell Biol.

    (1993)
  • M.S. Diamond et al.

    Heparin is an adhesive ligand for the leukocyte integrin Mac-1 (CD11b/CD18)

    J. Cell Biol.

    (1995)
  • I.C. Diller et al.

    The effect of yeast polysaccharides on mouse tumors

    Cancer Res.

    (1963)
  • N.R. Di Luzio et al.

    Comparative tumor-inhibitory and anti-bacterial activity of soluble and particulate glucan

    Int. J. Cancer

    (1979)
  • M. Doita et al.

    Effect of soluble aminated β-1,3-d-polyglucose on human monocytes: stimulation of cytokine and prostaglandin E2 production but not antigen-presenting function

    J. Leukocyte Biol.

    (1991)
  • F. Esteban et al.

    MHC class I antigens and tumour-infiltrating leucocytes in laryngeal cancer: long-term follow-up

    Br. J. Cancer

    (1996)
  • C.B. Forsyth et al.

    Interaction of the fungal pathogen Candida albicans with integrin CD11b/CD18: recognition by the I domain is modulated by the lectin-like domain and the CD18 subunit

    J. Immunol.

    (1998)
  • S. Fujimoto

    Clinical efficacies of schizophyllan (SPG) on advanced gastric cancer

    Nippon Geka Gakkai Zasshi

    (1989)
  • S. Fujimoto et al.

    Clinical evaluation of schizophyllan adjuvant immunochemotherapy for patients with resectable gastric cancer—a randomized controlled trial

    Jpn. J. Surg.

    (1984)
  • M.R. Gyetko et al.

    Function of the urokinase receptor (CD87) in neutrophil chemotaxis

    J. Leukocyte Biol.

    (1995)
  • J. Hamuro et al.

    Beta (1–3) glucans as a probe for T cell specific immune adjuvants: II. Enhanced in vitro generation of cytotoxic T lymphocytes

    Cell. Immunol.

    (1978)
  • D.J. Hicklin et al.

    β2-microglobulin mutations, HLA class I antigen loss, and tumor progression in melanoma

    J. Clin. Invest.

    (1998)
  • R.R. Ingalls et al.

    The CD11/CD18 integrins: characterization of three novel LPS signaling receptors

    Prog. Clin. Biol. Res.

    (1998)
  • K. Irie et al.

    Evidence for in vivo reaction of antibody and complement to surface antigens of human cancer cells

    Science

    (1974)
  • U.S. Kammula et al.

    Trends in the safety of high dose bolus interleukin-2 administration in patients with metastatic cancer

    Cancer

    (1998)
  • Cited by (240)

    View all citing articles on Scopus
    View full text